[petsc-dev] MatMatSolve and mumps

Jed Brown jedbrown at mcs.anl.gov
Wed Oct 30 20:51:03 CDT 2013


Jose David Bermeol <jbermeol at purdue.edu> writes:

> So for which solvers MatMatSolve does not work??

There is a default implementation (MatMatSolve_Basic) that just solves
single vector problems in sequence.  So you can use the interface, but
the execution might vary between solvers.

The MUMPS interface is actually a liability here because they insist on
having the entire vector on rank 0.  This is a terrible interface and
not memory-scalable.  If you naively put many RHS vectors on rank 0, you
would likely run out of memory.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: not available
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20131030/27381edb/attachment.sig>


More information about the petsc-dev mailing list